JSON-XS/XS.xs

#include "EXTERN.h"
#include "perl.h"
#include "XSUB.h"

#include "assert.h"
#include "string.h"
#include "stdlib.h"

#define F_ASCII        0x00000001
#define F_UTF8         0x00000002
#define F_INDENT       0x00000004
#define F_CANONICAL    0x00000008
#define F_SPACE_BEFORE 0x00000010
#define F_SPACE_AFTER  0x00000020
#define F_JSON_RPC     0x00000040
#define F_ALLOW_NONREF 0x00000080

#define F_PRETTY    F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER
#define F_DEFAULT   0

#define INIT_SIZE   32 // initial scalar size to be allocated

#define SB do {
#define SE } while (0)

static HV *json_stash;

// structure used for encoding JSON
typedef struct
{
  char *cur;
  STRLEN len; // SvLEN (sv)
  char *end;  // SvEND (sv)
  SV *sv;
  UV flags;
  int max_recurse;
  int indent;
} enc_t;

// structure used for decoding JSON
typedef struct
{
  char *cur;
  char *end;
  const char *err;
  UV flags;
} dec_t;

static UV *
SvJSON (SV *sv)
{
  if (!(SvROK (sv) && SvOBJECT (SvRV (sv)) && SvSTASH (SvRV (sv)) == json_stash))
    croak ("object is not of type JSON::XS");

  return &SvUVX (SvRV (sv));
}

/////////////////////////////////////////////////////////////////////////////

static void
need (enc_t *enc, STRLEN len)
{
  if (enc->cur + len >= enc->end)
    {
      STRLEN cur = enc->cur - SvPVX (enc->sv);
      SvGROW (enc->sv, cur + len + 1);
      enc->cur = SvPVX (enc->sv) + cur;
      enc->end = SvPVX (enc->sv) + SvLEN (enc->sv);
    }
}

static void
encode_ch (enc_t *enc, char ch)
{
  need (enc, 1);
  *enc->cur++ = ch;
}

static void
encode_str (enc_t *enc, char *str, STRLEN len, int is_utf8)
{
  char *end = str + len;

  need (enc, len);

  while (str < end)
    {
      unsigned char ch = *(unsigned char *)str;

      if (ch == '"')
        {
          need (enc, len += 1);
          *enc->cur++ = '\\';
          *enc->cur++ = '"';
          ++str;
        }
      else if (ch == '\\')
        {
          need (enc, len += 1);
          *enc->cur++ = '\\';
          *enc->cur++ = '\\';
          ++str;
        }
      else if (ch >= 0x20 && ch < 0x80) // most common case
        {
          *enc->cur++ = ch;
          ++str;
        }
      else if (ch == '\015')
        {
          need (enc, len += 1);
          *enc->cur++ = '\\';
          *enc->cur++ = 'r';
          ++str;
        }
      else if (ch == '\012')
        {
          need (enc, len += 1);
          *enc->cur++ = '\\';
          *enc->cur++ = 'n';
          ++str;
        }
      else
        {
          STRLEN clen;
          UV uch;

          if (is_utf8)
            {
              uch = utf8n_to_uvuni (str, end - str, &clen, UTF8_CHECK_ONLY);
              if (clen < 0)
                croak ("malformed UTF-8 character in string, cannot convert to JSON");
            }
          else
            {
              uch = ch;
              clen = 1;
            }

          if (uch < 0x80 || enc->flags & F_ASCII)
            {
              if (uch > 0xFFFFUL)
                {
                  need (enc, len += 11);
                  sprintf (enc->cur, "\\u%04x\\u%04x",
                           (uch - 0x10000) / 0x400 + 0xD800,
                           (uch - 0x10000) % 0x400 + 0xDC00);
                  enc->cur += 12;
                }
              else
                {
                  static char hexdigit [16] = "0123456789abcdef";
                  need (enc, len += 5);
                  *enc->cur++ = '\\';
                  *enc->cur++ = 'u';
                  *enc->cur++ = hexdigit [ uch >> 12      ];
                  *enc->cur++ = hexdigit [(uch >>  8) & 15];
                  *enc->cur++ = hexdigit [(uch >>  4) & 15];
                  *enc->cur++ = hexdigit [(uch >>  0) & 15];
                }

              str += clen;
            }
          else if (is_utf8)
            {
              need (enc, len += clen);
              while (clen--)
                *enc->cur++ = *str++;
            }
          else
            {
              need (enc, 10); // never more than 11 bytes needed
              enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);
              ++str;
            }
        }

      --len;
    }
}

#define INDENT SB \
  if (enc->flags & F_INDENT)            \
    {                                   \
      int i_;                           \
      need (enc, enc->indent);          \
      for (i_ = enc->indent * 3; i_--; )\
        encode_ch (enc, ' ');           \
    }                                   \
  SE

#define SPACE SB need (enc, 1); encode_ch (enc, ' '); SE
#define NL    SB if (enc->flags & F_INDENT) { need (enc, 1); encode_ch (enc, '\n'); } SE
#define COMMA SB \
  encode_ch (enc, ',');                 \
  if (enc->flags & F_INDENT)            \
    NL;                                 \
  else if (enc->flags & F_SPACE_AFTER)  \
    SPACE;                              \
  SE

static void encode_sv (enc_t *enc, SV *sv);

static void
encode_av (enc_t *enc, AV *av)
{
  int i, len = av_len (av);

  encode_ch (enc, '['); NL;
  ++enc->indent;

  for (i = 0; i <= len; ++i)
    {
      INDENT;
      encode_sv (enc, *av_fetch (av, i, 0));

      if (i < len)
        COMMA;
    }

  NL;

  --enc->indent;
  INDENT; encode_ch (enc, ']');
}

static void
encode_he (enc_t *enc, HE *he)
{
  encode_ch (enc, '"');

  if (HeKLEN (he) == HEf_SVKEY)
    {
      SV *sv = HeSVKEY (he);
      STRLEN len;
      char *str;
      
      SvGETMAGIC (sv);
      str = SvPV (sv, len);

      encode_str (enc, str, len, SvUTF8 (sv));
    }
  else
    encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));

  encode_ch (enc, '"');

  if (enc->flags & F_SPACE_BEFORE) SPACE;
  encode_ch (enc, ':');
  if (enc->flags & F_SPACE_AFTER ) SPACE;
  encode_sv (enc, HeVAL (he));
}

// compare hash entries, used when all keys are bytestrings
static int
he_cmp_fast (const void *a_, const void *b_)
{
  int cmp;

  HE *a = *(HE **)a_;
  HE *b = *(HE **)b_;

  STRLEN la = HeKLEN (a);
  STRLEN lb = HeKLEN (b);

  if (!(cmp == memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb)))
    cmp = la < lb ? -1 : la == lb ? 0 : 1;

  return cmp;
}

// compare hash entries, used when some keys are sv's or utf-x
static int
he_cmp_slow (const void *a, const void *b)
{
  return sv_cmp (HeSVKEY_force (*(HE **)a), HeSVKEY_force (*(HE **)b));
}

static void
encode_hv (enc_t *enc, HV *hv)
{
  int count, i;

  encode_ch (enc, '{'); NL; ++enc->indent;

  if ((count = hv_iterinit (hv)))
    {
      // for canonical output we have to sort by keys first
      // actually, this is mostly due to the stupid so-called
      // security workaround added somewhere in 5.8.x.
      // that randomises hash orderings
      if (enc->flags & F_CANONICAL)
        {
          HE *he, *hes [count];
          int fast = 1;

          i = 0;
          while ((he = hv_iternext (hv)))
            {
              hes [i++] = he;
              if (HeKLEN (he) < 0 || HeKUTF8 (he))
                fast = 0;
            }

          assert (i == count);

          if (fast)
            qsort (hes, count, sizeof (HE *), he_cmp_fast);
          else
            {
              // hack to disable "use bytes"
              COP *oldcop = PL_curcop, cop;
              cop.op_private = 0;
              PL_curcop = &cop;

              SAVETMPS;
              qsort (hes, count, sizeof (HE *), he_cmp_slow);
              FREETMPS;

              PL_curcop = oldcop;
            }

          for (i = 0; i < count; ++i)
            {
              INDENT;
              encode_he (enc, hes [i]);

              if (i < count - 1)
                COMMA;
            }

          NL;
        }
      else
        {
          SV *sv;
          HE *he = hv_iternext (hv);

          for (;;)
            {
              INDENT;
              encode_he (enc, he);

              if (!(he = hv_iternext (hv)))
                break;

              COMMA;
            }

          NL;
        }
    }

  --enc->indent; INDENT; encode_ch (enc, '}');
}

static void
encode_sv (enc_t *enc, SV *sv)
{
  SvGETMAGIC (sv);

  if (SvPOKp (sv))
    {
      STRLEN len;
      char *str = SvPV (sv, len);
      encode_ch (enc, '"');
      encode_str (enc, str, len, SvUTF8 (sv));
      encode_ch (enc, '"');
    }
  else if (SvNOKp (sv))
    {
      need (enc, NV_DIG + 32);
      Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);
      enc->cur += strlen (enc->cur);
    }
  else if (SvIOKp (sv))
    {
      need (enc, 64);
      enc->cur += 
         SvIsUV(sv)
            ? snprintf (enc->cur, 64, "%"UVuf, (UV)SvUVX (sv))
            : snprintf (enc->cur, 64, "%"IVdf, (IV)SvIVX (sv));
    }
  else if (SvROK (sv))
    {
      if (!--enc->max_recurse)
        croak ("data structure too deep (hit recursion limit)");

      sv = SvRV (sv);

      switch (SvTYPE (sv))
        {
          case SVt_PVAV: encode_av (enc, (AV *)sv); break;
          case SVt_PVHV: encode_hv (enc, (HV *)sv); break;

          default:
            croak ("JSON can only represent references to arrays or hashes");
        }
    }
  else if (!SvOK (sv))
    encode_str (enc, "null", 4, 0);
  else
    croak ("encountered perl type that JSON cannot handle");
}

static SV *
encode_json (SV *scalar, UV flags)
{
  if (!(flags & F_ALLOW_NONREF) && !SvROK (scalar))
    croak ("hash- or arraref required (not a simple scalar, use allow_nonref to allow this)");

  enc_t enc;
  enc.flags       = flags;
  enc.sv          = sv_2mortal (NEWSV (0, INIT_SIZE));
  enc.cur         = SvPVX (enc.sv);
  enc.end         = SvEND (enc.sv);
  enc.max_recurse = 0;
  enc.indent      = 0;

  SvPOK_only (enc.sv);
  encode_sv (&enc, scalar);

  if (!(flags & (F_ASCII | F_UTF8)))
    SvUTF8_on (enc.sv);

  SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
  return enc.sv;
}

/////////////////////////////////////////////////////////////////////////////

#define WS \
  for (;;)                              \
    {                                   \
      char ch = *dec->cur;              \
      if (ch > 0x20                     \
          || (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)) \
        break;                          \
      ++dec->cur;                       \
    }

#define ERR(reason) SB dec->err = reason; goto fail; SE
#define EXPECT_CH(ch) SB \
  if (*dec->cur != ch)          \
    ERR (# ch " expected");     \
  ++dec->cur;                   \
  SE

static SV *decode_sv (dec_t *dec);

static signed char decode_hexdigit[256];

static UV
decode_4hex (dec_t *dec)
{
  signed char d1, d2, d3, d4;

  d1 = decode_hexdigit [((unsigned char *)dec->cur) [0]];
  if (d1 < 0) ERR ("four hexadecimal digits expected");
  d2 = decode_hexdigit [((unsigned char *)dec->cur) [1]];
  if (d2 < 0) ERR ("four hexadecimal digits expected");
  d3 = decode_hexdigit [((unsigned char *)dec->cur) [2]];
  if (d3 < 0) ERR ("four hexadecimal digits expected");
  d4 = decode_hexdigit [((unsigned char *)dec->cur) [3]];
  if (d4 < 0) ERR ("four hexadecimal digits expected");

  dec->cur += 4;

  return ((UV)d1) << 12
       | ((UV)d2) <<  8
       | ((UV)d3) <<  4
       | ((UV)d4);

fail:
  return (UV)-1;
}

#define APPEND_GROW(n) SB \
  if (cur + (n) >= end)                         \
    {                                           \
      STRLEN ofs = cur - SvPVX (sv);            \
      SvGROW (sv, ofs + (n) + 1);               \
      cur = SvPVX (sv) + ofs;                   \
      end = SvEND (sv);                         \
    }                                           \
  SE

#define APPEND_CH(ch) SB \
  APPEND_GROW (1);      \
  *cur++ = (ch);        \
  SE

static SV *
decode_str (dec_t *dec)
{
  SV *sv = NEWSV (0,2);
  int utf8 = 0;
  char *cur = SvPVX (sv);
  char *end = SvEND (sv);

  for (;;)
    {
      unsigned char ch = *(unsigned char *)dec->cur;

      if (ch == '"')
        break;
      else if (ch == '\\')
        {
          switch (*++dec->cur)
            {
              case '\\':
              case '/':
              case '"': APPEND_CH (*dec->cur++); break;

              case 'b': APPEND_CH ('\010'); ++dec->cur; break;
              case 't': APPEND_CH ('\011'); ++dec->cur; break;
              case 'n': APPEND_CH ('\012'); ++dec->cur; break;
              case 'f': APPEND_CH ('\014'); ++dec->cur; break;
              case 'r': APPEND_CH ('\015'); ++dec->cur; break;

              case 'u':
                {
                  UV lo, hi;
                  ++dec->cur;

                  hi = decode_4hex (dec);
                  if (hi == (UV)-1)
                    goto fail;

                  // possibly a surrogate pair
                  if (hi >= 0xd800 && hi < 0xdc00)
                    {
                      if (dec->cur [0] != '\\' || dec->cur [1] != 'u')
                        ERR ("illegal surrogate character");

                      dec->cur += 2;

                      lo = decode_4hex (dec);
                      if (lo == (UV)-1)
                        goto fail;

                      if (lo < 0xdc00 || lo >= 0xe000)
                        ERR ("surrogate pair expected");

                      hi = (hi - 0xD800) * 0x400 + (lo - 0xDC00) + 0x10000;
                    }
                  else if (lo >= 0xdc00 && lo < 0xe000)
                    ERR ("illegal surrogate character");

                  if (hi >= 0x80)
                    {
                      utf8 = 1;

                      APPEND_GROW (4); // at most 4 bytes for 21 bits
                      cur = (char *)uvuni_to_utf8_flags (cur, hi, 0);
                    }
                  else
                    APPEND_CH (hi);
                }
                break;
            }
        }
      else if (ch >= 0x20 && ch <= 0x7f)
        APPEND_CH (*dec->cur++);
      else if (ch >= 0x80)
        {
          STRLEN clen;
          UV uch = utf8n_to_uvuni (dec->cur, dec->end - dec->cur, &clen, UTF8_CHECK_ONLY);
          if (clen < 0)
            ERR ("malformed UTF-8 character in string, cannot convert to JSON");

          APPEND_GROW (clen);
          memcpy (cur, dec->cur, clen);
          cur += clen;
          dec->cur += clen;
        }
      else
        ERR ("invalid character encountered");
    }

  ++dec->cur;

  SvCUR_set (sv, cur - SvPVX (sv));

  SvPOK_only (sv);
  *SvEND (sv) = 0;

  if (utf8)
    SvUTF8_on (sv);

  return sv;

fail:
  SvREFCNT_dec (sv);
  return 0;
}

static SV *
decode_num (dec_t *dec)
{
  int is_nv = 0;
  char *start = dec->cur;

  // [minus]
  if (*dec->cur == '-')
    ++dec->cur;

  if (*dec->cur == '0')
    {
      ++dec->cur;
      if (*dec->cur >= '0' && *dec->cur <= '9')
         ERR ("malformed number (leading zero must not be followed by another digit)");
    }

  // int
  while (*dec->cur >= '0' && *dec->cur <= '9')
    ++dec->cur;

  // [frac]
  if (*dec->cur == '.')
    {
      is_nv = 1;

      do
        {
          ++dec->cur;
        }
      while (*dec->cur >= '0' && *dec->cur <= '9');
    }

  // [exp]
  if (*dec->cur == 'e' || *dec->cur == 'E')
    {
      is_nv = 1;

      ++dec->cur;
      if (*dec->cur == '-' || *dec->cur == '+')
        ++dec->cur;

      while (*dec->cur >= '0' && *dec->cur <= '9')
        ++dec->cur;
    }

  if (!is_nv)
    {
      UV uv;
      int numtype = grok_number (start, dec->cur - start, &uv);
      if (numtype & IS_NUMBER_IN_UV)
        if (numtype & IS_NUMBER_NEG)
          {
            if (uv < (UV)IV_MIN)
              return newSViv (-(IV)uv);
          }
        else
          return newSVuv (uv);
    }

  return newSVnv (Atof (start));

fail:
  return 0;
}

static SV *
decode_av (dec_t *dec)
{
  AV *av = newAV ();

  for (;;)
    {
      SV *value;

      value = decode_sv (dec);
      if (!value)
        goto fail;

      av_push (av, value);

      WS;

      if (*dec->cur == ']')
        {
          ++dec->cur;
          break;
        }
      
      if (*dec->cur != ',')
        ERR (", or ] expected while parsing array");

      ++dec->cur;
    }

  return newRV_noinc ((SV *)av);

fail:
  SvREFCNT_dec (av);
  return 0;
}

static SV *
decode_hv (dec_t *dec)
{
  HV *hv = newHV ();

  for (;;)
    {
      SV *key, *value;

      WS; EXPECT_CH ('"');

      key = decode_str (dec);
      if (!key)
        goto fail;

      WS; EXPECT_CH (':');

      value = decode_sv (dec);
      if (!value)
        {
          SvREFCNT_dec (key);
          goto fail;
        }

      //TODO: optimise
      hv_store_ent (hv, key, value, 0);

      WS;

      if (*dec->cur == '}')
        {
          ++dec->cur;
          break;
        }

      if (*dec->cur != ',')
        ERR (", or } expected while parsing object/hash");

      ++dec->cur;
    }

  return newRV_noinc ((SV *)hv);

fail:
  SvREFCNT_dec (hv);
  return 0;
}

static SV *
decode_sv (dec_t *dec)
{
  WS;
  switch (*dec->cur)
    {
      case '"': ++dec->cur; return decode_str (dec); 
      case '[': ++dec->cur; return decode_av (dec); 
      case '{': ++dec->cur; return decode_hv (dec);

      case '-':
      case '0': case '1': case '2': case '3': case '4':
      case '5': case '6': case '7': case '8': case '9':
        return decode_num (dec);

      case 't':
        if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
          {
            dec->cur += 4;
            return newSViv (1);
          }
        else
          ERR ("'true' expected");

        break;

      case 'f':
        if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
          {
            dec->cur += 5;
            return newSViv (0);
          }
        else
          ERR ("'false' expected");

        break;

      case 'n':
        if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "null", 4))
          {
            dec->cur += 4;
            return newSViv (1);
          }
        else
          ERR ("'null' expected");

        break;

      default:
        ERR ("malformed json string");
        break;
    }

fail:
  return 0;
}

static SV *
decode_json (SV *string, UV flags)
{
  SV *sv;

  if (!(flags & F_UTF8))
    sv_utf8_upgrade (string);

  SvGROW (string, SvCUR (string) + 1); // should basically be a NOP

  dec_t dec;
  dec.flags = flags;
  dec.cur   = SvPVX (string);
  dec.end   = SvEND (string);
  dec.err   = 0;

  *dec.end = 1; // invalid anywhere
  sv = decode_sv (&dec);
  *dec.end = 0;

  if (!sv)
    {
      IV offset = utf8_distance (dec.cur, SvPVX (string));
      SV *uni = sv_newmortal ();

      pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);
      croak ("%s, at character %d (%s)",
             dec.err,
             (int)offset,
             dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
    }

  sv = sv_2mortal (sv);

  if (!(dec.flags & F_ALLOW_NONREF) && !SvROK (sv))
    croak ("JSON object or array expected (but number, string, true, false or null found, use allow_nonref to allow this)");

  return sv;
}

MODULE = JSON::XS               PACKAGE = JSON::XS

BOOT:
{
        int i;

        memset (decode_hexdigit, 0xff, 256);
        for (i = 10; i--; )
          decode_hexdigit ['0' + i] = i;

        for (i = 7; i--; )
          {
            decode_hexdigit ['a' + i] = 10 + i;
            decode_hexdigit ['A' + i] = 10 + i;
          }

        json_stash = gv_stashpv ("JSON::XS", 1);
}

PROTOTYPES: DISABLE

SV *new (char *dummy)
        CODE:
        RETVAL = sv_bless (newRV_noinc (newSVuv (F_DEFAULT)), json_stash);
        OUTPUT:
        RETVAL

SV *ascii (SV *self, int enable)
        ALIAS:
        ascii        = F_ASCII
        utf8         = F_UTF8
        indent       = F_INDENT
        canonical    = F_CANONICAL
        space_before = F_SPACE_BEFORE
        space_after  = F_SPACE_AFTER
        json_rpc     = F_JSON_RPC
        pretty       = F_PRETTY
        allow_nonref = F_ALLOW_NONREF
        CODE:
{
        UV *uv = SvJSON (self);
        if (enable)
          *uv |=  ix;
        else
          *uv &= ~ix;

        RETVAL = newSVsv (self);
}
        OUTPUT:
        RETVAL

void encode (SV *self, SV *scalar)
        PPCODE:
        XPUSHs (encode_json (scalar, *SvJSON (self)));

void decode (SV *self, SV *jsonstr)
        PPCODE:
        XPUSHs (decode_json (jsonstr, *SvJSON (self)));

PROTOTYPES: ENABLE

void to_json (SV *scalar)
        PPCODE:
        XPUSHs (encode_json (scalar, F_UTF8));

void from_json (SV *jsonstr)
        PPCODE:
        XPUSHs (decode_json (jsonstr, F_UTF8));

Revision:	1.4
Committed:	Thu Mar 22 21:13:58 2007 UTC (17 years, 2 months ago) by root
Branch:	MAIN
CVS Tags:	rel-0_1
Changes since 1.3:	+90 -30 lines
Log Message:	* empty log message *
#	Content
1	#include "EXTERN.h"
2	#include "perl.h"
3	#include "XSUB.h"
4
5	#include "assert.h"
6	#include "string.h"
7	#include "stdlib.h"
8
9	#define F_ASCII 0x00000001
10	#define F_UTF8 0x00000002
11	#define F_INDENT 0x00000004
12	#define F_CANONICAL 0x00000008
13	#define F_SPACE_BEFORE 0x00000010
14	#define F_SPACE_AFTER 0x00000020
15	#define F_JSON_RPC 0x00000040
16	#define F_ALLOW_NONREF 0x00000080
17
18	#define F_PRETTY F_INDENT \| F_SPACE_BEFORE \| F_SPACE_AFTER
19	#define F_DEFAULT 0
20
21	#define INIT_SIZE 32 // initial scalar size to be allocated
22
23	#define SB do {
24	#define SE } while (0)
25
26	static HV *json_stash;
27
28	// structure used for encoding JSON
29	typedef struct
30	{
31	char *cur;
32	STRLEN len; // SvLEN (sv)
33	char *end; // SvEND (sv)
34	SV *sv;
35	UV flags;
36	int max_recurse;
37	int indent;
38	} enc_t;
39
40	// structure used for decoding JSON
41	typedef struct
42	{
43	char *cur;
44	char *end;
45	const char *err;
46	UV flags;
47	} dec_t;
48
49	static UV *
50	SvJSON (SV *sv)
51	{
52	if (!(SvROK (sv) && SvOBJECT (SvRV (sv)) && SvSTASH (SvRV (sv)) == json_stash))
53	croak ("object is not of type JSON::XS");
54
55	return &SvUVX (SvRV (sv));
56	}
57
58	/////////////////////////////////////////////////////////////////////////////
59
60	static void
61	need (enc_t *enc, STRLEN len)
62	{
63	if (enc->cur + len >= enc->end)
64	{
65	STRLEN cur = enc->cur - SvPVX (enc->sv);
66	SvGROW (enc->sv, cur + len + 1);
67	enc->cur = SvPVX (enc->sv) + cur;
68	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv);
69	}
70	}
71
72	static void
73	encode_ch (enc_t *enc, char ch)
74	{
75	need (enc, 1);
76	*enc->cur++ = ch;
77	}
78
79	static void
80	encode_str (enc_t enc, char str, STRLEN len, int is_utf8)
81	{
82	char *end = str + len;
83
84	need (enc, len);
85
86	while (str < end)
87	{
88	unsigned char ch = (unsigned char )str;
89
90	if (ch == '"')
91	{
92	need (enc, len += 1);
93	*enc->cur++ = '\\';
94	*enc->cur++ = '"';
95	++str;
96	}
97	else if (ch == '\\')
98	{
99	need (enc, len += 1);
100	*enc->cur++ = '\\';
101	*enc->cur++ = '\\';
102	++str;
103	}
104	else if (ch >= 0x20 && ch < 0x80) // most common case
105	{
106	*enc->cur++ = ch;
107	++str;
108	}
109	else if (ch == '\015')
110	{
111	need (enc, len += 1);
112	*enc->cur++ = '\\';
113	*enc->cur++ = 'r';
114	++str;
115	}
116	else if (ch == '\012')
117	{
118	need (enc, len += 1);
119	*enc->cur++ = '\\';
120	*enc->cur++ = 'n';
121	++str;
122	}
123	else
124	{
125	STRLEN clen;
126	UV uch;
127
128	if (is_utf8)
129	{
130	uch = utf8n_to_uvuni (str, end - str, &clen, UTF8_CHECK_ONLY);
131	if (clen < 0)
132	croak ("malformed UTF-8 character in string, cannot convert to JSON");
133	}
134	else
135	{
136	uch = ch;
137	clen = 1;
138	}
139
140	if (uch < 0x80 \|\| enc->flags & F_ASCII)
141	{
142	if (uch > 0xFFFFUL)
143	{
144	need (enc, len += 11);
145	sprintf (enc->cur, "\\u%04x\\u%04x",
146	(uch - 0x10000) / 0x400 + 0xD800,
147	(uch - 0x10000) % 0x400 + 0xDC00);
148	enc->cur += 12;
149	}
150	else
151	{
152	static char hexdigit [16] = "0123456789abcdef";
153	need (enc, len += 5);
154	*enc->cur++ = '\\';
155	*enc->cur++ = 'u';
156	*enc->cur++ = hexdigit [ uch >> 12 ];
157	*enc->cur++ = hexdigit [(uch >> 8) & 15];
158	*enc->cur++ = hexdigit [(uch >> 4) & 15];
159	*enc->cur++ = hexdigit [(uch >> 0) & 15];
160	}
161
162	str += clen;
163	}
164	else if (is_utf8)
165	{
166	need (enc, len += clen);
167	while (clen--)
168	enc->cur++ = str++;
169	}
170	else
171	{
172	need (enc, 10); // never more than 11 bytes needed
173	enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);
174	++str;
175	}
176	}
177
178	--len;
179	}
180	}
181
182	#define INDENT SB \
183	if (enc->flags & F_INDENT) \
184	{ \
185	int i_; \
186	need (enc, enc->indent); \
187	for (i_ = enc->indent * 3; i_--; )\
188	encode_ch (enc, ' '); \
189	} \
190	SE
191
192	#define SPACE SB need (enc, 1); encode_ch (enc, ' '); SE
193	#define NL SB if (enc->flags & F_INDENT) { need (enc, 1); encode_ch (enc, '\n'); } SE
194	#define COMMA SB \
195	encode_ch (enc, ','); \
196	if (enc->flags & F_INDENT) \
197	NL; \
198	else if (enc->flags & F_SPACE_AFTER) \
199	SPACE; \
200	SE
201
202	static void encode_sv (enc_t enc, SV sv);
203
204	static void
205	encode_av (enc_t enc, AV av)
206	{
207	int i, len = av_len (av);
208
209	encode_ch (enc, '['); NL;
210	++enc->indent;
211
212	for (i = 0; i <= len; ++i)
213	{
214	INDENT;
215	encode_sv (enc, *av_fetch (av, i, 0));
216
217	if (i < len)
218	COMMA;
219	}
220
221	NL;
222
223	--enc->indent;
224	INDENT; encode_ch (enc, ']');
225	}
226
227	static void
228	encode_he (enc_t enc, HE he)
229	{
230	encode_ch (enc, '"');
231
232	if (HeKLEN (he) == HEf_SVKEY)
233	{
234	SV *sv = HeSVKEY (he);
235	STRLEN len;
236	char *str;
237
238	SvGETMAGIC (sv);
239	str = SvPV (sv, len);
240
241	encode_str (enc, str, len, SvUTF8 (sv));
242	}
243	else
244	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));
245
246	encode_ch (enc, '"');
247
248	if (enc->flags & F_SPACE_BEFORE) SPACE;
249	encode_ch (enc, ':');
250	if (enc->flags & F_SPACE_AFTER ) SPACE;
251	encode_sv (enc, HeVAL (he));
252	}
253
254	// compare hash entries, used when all keys are bytestrings
255	static int
256	he_cmp_fast (const void a_, const void b_)
257	{
258	int cmp;
259
260	HE a = (HE **)a_;
261	HE b = (HE **)b_;
262
263	STRLEN la = HeKLEN (a);
264	STRLEN lb = HeKLEN (b);
265
266	if (!(cmp == memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb)))
267	cmp = la < lb ? -1 : la == lb ? 0 : 1;
268
269	return cmp;
270	}
271
272	// compare hash entries, used when some keys are sv's or utf-x
273	static int
274	he_cmp_slow (const void a, const void b)
275	{
276	return sv_cmp (HeSVKEY_force ((HE )a), HeSVKEY_force ((HE **)b));
277	}
278
279	static void
280	encode_hv (enc_t enc, HV hv)
281	{
282	int count, i;
283
284	encode_ch (enc, '{'); NL; ++enc->indent;
285
286	if ((count = hv_iterinit (hv)))
287	{
288	// for canonical output we have to sort by keys first
289	// actually, this is mostly due to the stupid so-called
290	// security workaround added somewhere in 5.8.x.
291	// that randomises hash orderings
292	if (enc->flags & F_CANONICAL)
293	{
294	HE he, hes [count];
295	int fast = 1;
296
297	i = 0;
298	while ((he = hv_iternext (hv)))
299	{
300	hes [i++] = he;
301	if (HeKLEN (he) < 0 \|\| HeKUTF8 (he))
302	fast = 0;
303	}
304
305	assert (i == count);
306
307	if (fast)
308	qsort (hes, count, sizeof (HE *), he_cmp_fast);
309	else
310	{
311	// hack to disable "use bytes"
312	COP *oldcop = PL_curcop, cop;
313	cop.op_private = 0;
314	PL_curcop = &cop;
315
316	SAVETMPS;
317	qsort (hes, count, sizeof (HE *), he_cmp_slow);
318	FREETMPS;
319
320	PL_curcop = oldcop;
321	}
322
323	for (i = 0; i < count; ++i)
324	{
325	INDENT;
326	encode_he (enc, hes [i]);
327
328	if (i < count - 1)
329	COMMA;
330	}
331
332	NL;
333	}
334	else
335	{
336	SV *sv;
337	HE *he = hv_iternext (hv);
338
339	for (;;)
340	{
341	INDENT;
342	encode_he (enc, he);
343
344	if (!(he = hv_iternext (hv)))
345	break;
346
347	COMMA;
348	}
349
350	NL;
351	}
352	}
353
354	--enc->indent; INDENT; encode_ch (enc, '}');
355	}
356
357	static void
358	encode_sv (enc_t enc, SV sv)
359	{
360	SvGETMAGIC (sv);
361
362	if (SvPOKp (sv))
363	{
364	STRLEN len;
365	char *str = SvPV (sv, len);
366	encode_ch (enc, '"');
367	encode_str (enc, str, len, SvUTF8 (sv));
368	encode_ch (enc, '"');
369	}
370	else if (SvNOKp (sv))
371	{
372	need (enc, NV_DIG + 32);
373	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);
374	enc->cur += strlen (enc->cur);
375	}
376	else if (SvIOKp (sv))
377	{
378	need (enc, 64);
379	enc->cur +=
380	SvIsUV(sv)
381	? snprintf (enc->cur, 64, "%"UVuf, (UV)SvUVX (sv))
382	: snprintf (enc->cur, 64, "%"IVdf, (IV)SvIVX (sv));
383	}
384	else if (SvROK (sv))
385	{
386	if (!--enc->max_recurse)
387	croak ("data structure too deep (hit recursion limit)");
388
389	sv = SvRV (sv);
390
391	switch (SvTYPE (sv))
392	{
393	case SVt_PVAV: encode_av (enc, (AV *)sv); break;
394	case SVt_PVHV: encode_hv (enc, (HV *)sv); break;
395
396	default:
397	croak ("JSON can only represent references to arrays or hashes");
398	}
399	}
400	else if (!SvOK (sv))
401	encode_str (enc, "null", 4, 0);
402	else
403	croak ("encountered perl type that JSON cannot handle");
404	}
405
406	static SV *
407	encode_json (SV *scalar, UV flags)
408	{
409	if (!(flags & F_ALLOW_NONREF) && !SvROK (scalar))
410	croak ("hash- or arraref required (not a simple scalar, use allow_nonref to allow this)");
411
412	enc_t enc;
413	enc.flags = flags;
414	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
415	enc.cur = SvPVX (enc.sv);
416	enc.end = SvEND (enc.sv);
417	enc.max_recurse = 0;
418	enc.indent = 0;
419
420	SvPOK_only (enc.sv);
421	encode_sv (&enc, scalar);
422
423	if (!(flags & (F_ASCII \| F_UTF8)))
424	SvUTF8_on (enc.sv);
425
426	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
427	return enc.sv;
428	}
429
430	/////////////////////////////////////////////////////////////////////////////
431
432	#define WS \
433	for (;;) \
434	{ \
435	char ch = *dec->cur; \
436	if (ch > 0x20 \
437	\|\| (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)) \
438	break; \
439	++dec->cur; \
440	}
441
442	#define ERR(reason) SB dec->err = reason; goto fail; SE
443	#define EXPECT_CH(ch) SB \
444	if (*dec->cur != ch) \
445	ERR (# ch " expected"); \
446	++dec->cur; \
447	SE
448
449	static SV decode_sv (dec_t dec);
450
451	static signed char decode_hexdigit[256];
452
453	static UV
454	decode_4hex (dec_t *dec)
455	{
456	signed char d1, d2, d3, d4;
457
458	d1 = decode_hexdigit [((unsigned char *)dec->cur) [0]];
459	if (d1 < 0) ERR ("four hexadecimal digits expected");
460	d2 = decode_hexdigit [((unsigned char *)dec->cur) [1]];
461	if (d2 < 0) ERR ("four hexadecimal digits expected");
462	d3 = decode_hexdigit [((unsigned char *)dec->cur) [2]];
463	if (d3 < 0) ERR ("four hexadecimal digits expected");
464	d4 = decode_hexdigit [((unsigned char *)dec->cur) [3]];
465	if (d4 < 0) ERR ("four hexadecimal digits expected");
466
467	dec->cur += 4;
468
469	return ((UV)d1) << 12
470	\| ((UV)d2) << 8
471	\| ((UV)d3) << 4
472	\| ((UV)d4);
473
474	fail:
475	return (UV)-1;
476	}
477
478	#define APPEND_GROW(n) SB \
479	if (cur + (n) >= end) \
480	{ \
481	STRLEN ofs = cur - SvPVX (sv); \
482	SvGROW (sv, ofs + (n) + 1); \
483	cur = SvPVX (sv) + ofs; \
484	end = SvEND (sv); \
485	} \
486	SE
487
488	#define APPEND_CH(ch) SB \
489	APPEND_GROW (1); \
490	*cur++ = (ch); \
491	SE
492
493	static SV *
494	decode_str (dec_t *dec)
495	{
496	SV *sv = NEWSV (0,2);
497	int utf8 = 0;
498	char *cur = SvPVX (sv);
499	char *end = SvEND (sv);
500
501	for (;;)
502	{
503	unsigned char ch = (unsigned char )dec->cur;
504
505	if (ch == '"')
506	break;
507	else if (ch == '\\')
508	{
509	switch (*++dec->cur)
510	{
511	case '\\':
512	case '/':
513	case '"': APPEND_CH (*dec->cur++); break;
514
515	case 'b': APPEND_CH ('\010'); ++dec->cur; break;
516	case 't': APPEND_CH ('\011'); ++dec->cur; break;
517	case 'n': APPEND_CH ('\012'); ++dec->cur; break;
518	case 'f': APPEND_CH ('\014'); ++dec->cur; break;
519	case 'r': APPEND_CH ('\015'); ++dec->cur; break;
520
521	case 'u':
522	{
523	UV lo, hi;
524	++dec->cur;
525
526	hi = decode_4hex (dec);
527	if (hi == (UV)-1)
528	goto fail;
529
530	// possibly a surrogate pair
531	if (hi >= 0xd800 && hi < 0xdc00)
532	{
533	if (dec->cur [0] != '\\' \|\| dec->cur [1] != 'u')
534	ERR ("illegal surrogate character");
535
536	dec->cur += 2;
537
538	lo = decode_4hex (dec);
539	if (lo == (UV)-1)
540	goto fail;
541
542	if (lo < 0xdc00 \|\| lo >= 0xe000)
543	ERR ("surrogate pair expected");
544
545	hi = (hi - 0xD800) * 0x400 + (lo - 0xDC00) + 0x10000;
546	}
547	else if (lo >= 0xdc00 && lo < 0xe000)
548	ERR ("illegal surrogate character");
549
550	if (hi >= 0x80)
551	{
552	utf8 = 1;
553
554	APPEND_GROW (4); // at most 4 bytes for 21 bits
555	cur = (char *)uvuni_to_utf8_flags (cur, hi, 0);
556	}
557	else
558	APPEND_CH (hi);
559	}
560	break;
561	}
562	}
563	else if (ch >= 0x20 && ch <= 0x7f)
564	APPEND_CH (*dec->cur++);
565	else if (ch >= 0x80)
566	{
567	STRLEN clen;
568	UV uch = utf8n_to_uvuni (dec->cur, dec->end - dec->cur, &clen, UTF8_CHECK_ONLY);
569	if (clen < 0)
570	ERR ("malformed UTF-8 character in string, cannot convert to JSON");
571
572	APPEND_GROW (clen);
573	memcpy (cur, dec->cur, clen);
574	cur += clen;
575	dec->cur += clen;
576	}
577	else
578	ERR ("invalid character encountered");
579	}
580
581	++dec->cur;
582
583	SvCUR_set (sv, cur - SvPVX (sv));
584
585	SvPOK_only (sv);
586	*SvEND (sv) = 0;
587
588	if (utf8)
589	SvUTF8_on (sv);
590
591	return sv;
592
593	fail:
594	SvREFCNT_dec (sv);
595	return 0;
596	}
597
598	static SV *
599	decode_num (dec_t *dec)
600	{
601	int is_nv = 0;
602	char *start = dec->cur;
603
604	// [minus]
605	if (*dec->cur == '-')
606	++dec->cur;
607
608	if (*dec->cur == '0')
609	{
610	++dec->cur;
611	if (dec->cur >= '0' && dec->cur <= '9')
612	ERR ("malformed number (leading zero must not be followed by another digit)");
613	}
614
615	// int
616	while (dec->cur >= '0' && dec->cur <= '9')
617	++dec->cur;
618
619	// [frac]
620	if (*dec->cur == '.')
621	{
622	is_nv = 1;
623
624	do
625	{
626	++dec->cur;
627	}
628	while (dec->cur >= '0' && dec->cur <= '9');
629	}
630
631	// [exp]
632	if (dec->cur == 'e' \|\| dec->cur == 'E')
633	{
634	is_nv = 1;
635
636	++dec->cur;
637	if (dec->cur == '-' \|\| dec->cur == '+')
638	++dec->cur;
639
640	while (dec->cur >= '0' && dec->cur <= '9')
641	++dec->cur;
642	}
643
644	if (!is_nv)
645	{
646	UV uv;
647	int numtype = grok_number (start, dec->cur - start, &uv);
648	if (numtype & IS_NUMBER_IN_UV)
649	if (numtype & IS_NUMBER_NEG)
650	{
651	if (uv < (UV)IV_MIN)
652	return newSViv (-(IV)uv);
653	}
654	else
655	return newSVuv (uv);
656	}
657
658	return newSVnv (Atof (start));
659
660	fail:
661	return 0;
662	}
663
664	static SV *
665	decode_av (dec_t *dec)
666	{
667	AV *av = newAV ();
668
669	for (;;)
670	{
671	SV *value;
672
673	value = decode_sv (dec);
674	if (!value)
675	goto fail;
676
677	av_push (av, value);
678
679	WS;
680
681	if (*dec->cur == ']')
682	{
683	++dec->cur;
684	break;
685	}
686
687	if (*dec->cur != ',')
688	ERR (", or ] expected while parsing array");
689
690	++dec->cur;
691	}
692
693	return newRV_noinc ((SV *)av);
694
695	fail:
696	SvREFCNT_dec (av);
697	return 0;
698	}
699
700	static SV *
701	decode_hv (dec_t *dec)
702	{
703	HV *hv = newHV ();
704
705	for (;;)
706	{
707	SV key, value;
708
709	WS; EXPECT_CH ('"');
710
711	key = decode_str (dec);
712	if (!key)
713	goto fail;
714
715	WS; EXPECT_CH (':');
716
717	value = decode_sv (dec);
718	if (!value)
719	{
720	SvREFCNT_dec (key);
721	goto fail;
722	}
723
724	//TODO: optimise
725	hv_store_ent (hv, key, value, 0);
726
727	WS;
728
729	if (*dec->cur == '}')
730	{
731	++dec->cur;
732	break;
733	}
734
735	if (*dec->cur != ',')
736	ERR (", or } expected while parsing object/hash");
737
738	++dec->cur;
739	}
740
741	return newRV_noinc ((SV *)hv);
742
743	fail:
744	SvREFCNT_dec (hv);
745	return 0;
746	}
747
748	static SV *
749	decode_sv (dec_t *dec)
750	{
751	WS;
752	switch (*dec->cur)
753	{
754	case '"': ++dec->cur; return decode_str (dec);
755	case '[': ++dec->cur; return decode_av (dec);
756	case '{': ++dec->cur; return decode_hv (dec);
757
758	case '-':
759	case '0': case '1': case '2': case '3': case '4':
760	case '5': case '6': case '7': case '8': case '9':
761	return decode_num (dec);
762
763	case 't':
764	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
765	{
766	dec->cur += 4;
767	return newSViv (1);
768	}
769	else
770	ERR ("'true' expected");
771
772	break;
773
774	case 'f':
775	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
776	{
777	dec->cur += 5;
778	return newSViv (0);
779	}
780	else
781	ERR ("'false' expected");
782
783	break;
784
785	case 'n':
786	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "null", 4))
787	{
788	dec->cur += 4;
789	return newSViv (1);
790	}
791	else
792	ERR ("'null' expected");
793
794	break;
795
796	default:
797	ERR ("malformed json string");
798	break;
799	}
800
801	fail:
802	return 0;
803	}
804
805	static SV *
806	decode_json (SV *string, UV flags)
807	{
808	SV *sv;
809
810	if (!(flags & F_UTF8))
811	sv_utf8_upgrade (string);
812
813	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
814
815	dec_t dec;
816	dec.flags = flags;
817	dec.cur = SvPVX (string);
818	dec.end = SvEND (string);
819	dec.err = 0;
820
821	*dec.end = 1; // invalid anywhere
822	sv = decode_sv (&dec);
823	*dec.end = 0;
824
825	if (!sv)
826	{
827	IV offset = utf8_distance (dec.cur, SvPVX (string));
828	SV *uni = sv_newmortal ();
829
830	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);
831	croak ("%s, at character %d (%s)",
832	dec.err,
833	(int)offset,
834	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
835	}
836
837	sv = sv_2mortal (sv);
838
839	if (!(dec.flags & F_ALLOW_NONREF) && !SvROK (sv))
840	croak ("JSON object or array expected (but number, string, true, false or null found, use allow_nonref to allow this)");
841
842	return sv;
843	}
844
845	MODULE = JSON::XS PACKAGE = JSON::XS
846
847	BOOT:
848	{
849	int i;
850
851	memset (decode_hexdigit, 0xff, 256);
852	for (i = 10; i--; )
853	decode_hexdigit ['0' + i] = i;
854
855	for (i = 7; i--; )
856	{
857	decode_hexdigit ['a' + i] = 10 + i;
858	decode_hexdigit ['A' + i] = 10 + i;
859	}
860
861	json_stash = gv_stashpv ("JSON::XS", 1);
862	}
863
864	PROTOTYPES: DISABLE
865
866	SV new (char dummy)
867	CODE:
868	RETVAL = sv_bless (newRV_noinc (newSVuv (F_DEFAULT)), json_stash);
869	OUTPUT:
870	RETVAL
871
872	SV ascii (SV self, int enable)
873	ALIAS:
874	ascii = F_ASCII
875	utf8 = F_UTF8
876	indent = F_INDENT
877	canonical = F_CANONICAL
878	space_before = F_SPACE_BEFORE
879	space_after = F_SPACE_AFTER
880	json_rpc = F_JSON_RPC
881	pretty = F_PRETTY
882	allow_nonref = F_ALLOW_NONREF
883	CODE:
884	{
885	UV *uv = SvJSON (self);
886	if (enable)
887	*uv \|= ix;
888	else
889	*uv &= ~ix;
890
891	RETVAL = newSVsv (self);
892	}
893	OUTPUT:
894	RETVAL
895
896	void encode (SV self, SV scalar)
897	PPCODE:
898	XPUSHs (encode_json (scalar, *SvJSON (self)));
899
900	void decode (SV self, SV jsonstr)
901	PPCODE:
902	XPUSHs (decode_json (jsonstr, *SvJSON (self)));
903
904	PROTOTYPES: ENABLE
905
906	void to_json (SV *scalar)
907	PPCODE:
908	XPUSHs (encode_json (scalar, F_UTF8));
909
910	void from_json (SV *jsonstr)
911	PPCODE:
912	XPUSHs (decode_json (jsonstr, F_UTF8));
913